Abstracts

To determine whether the spatial networks underlying generalised spike-wave activity during human absence seizures are identifiable using fMRI data alone. Spatiotemporal Independent Component Analysis (ICA) was performed using MELODIC (Multivariate Exploratory Linear Decomposition into Independent Components) version 2.0, part of the FSL software package (FMRIBs Software Library, http://www.fmrib.ox.ac.uk/fsl), on ictal fMRI data acquired previously from a patient with intractable Idiopathic Generalised Epilepsy¹.
Pre-processed data (masked, mean corrected and variance normalised) was whitened and projected into a 111-dimensional subspace using probabilistic Principal Component Analysis where the number of dimensions was estimated automatically. The observations were decomposed into a set of time-courses and spatial maps by optimising for non-Gaussian spatial source distributions using a fixed-point iteration technique. Estimated Component maps were divided by the standard deviation of the residual noise and thresholded by fitting a mixture model to the histogram of intensity values (probabilistic ICA). All time-courses were ranked according to degree of correlation with seizure onset and the corresponding spatial components contrasted with the results of a General Linear Model (GLM) analysis. At least ten independent components were identified as contributing to the previously reported pattern of thalamic activation and cortical deactivation, each with a timecourse mirroring the ictal EEG activity (see Figure 1).[figure1] - Five thresholded IC maps (alternative hypothesis test at p [gt] 0.5) are shown with their corresponding timecourses underneath. Four absence seizures are indicated with gray shading. Our observations indicate that a weighted mixture of several [underline]independent[/underline] spatial networks may simultaneously subserve the generation of generalised spike-wave activity in man. A mechanism is therefore provided, whereby subject or syndrome-specific changes in this blend, could account for a number of apparent discrepancies in the EEG/fMRI literature^1-3.
Our results also support the notion that the haemodynamic consequences of prolonged spike-wave activity may be identifiable using fMRI alone.
¹Salek-Haddadi et al. [italic]Annals of Neurology[/italic] 2003 ; 53(5) :663-667
²Archer et al [italic]Neuroimage[/italic] 2003; 20(4):1915-1922.
³Aghakhani et al. [italic]Brain[/italic] 2003 In Press. (Supported by Medical Research Council (UK) and UCL-CRDC)